SEARCH

SEARCH BY CITATION

Keywords:

  • Salmonella;
  • pigs;
  • zoonotic;
  • Denmark;
  • surveillance;
  • zero-inflated

Summary

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References

The national control programme for Salmonella in Danish swine herds introduced in 1993 has led to a large decrease in pork-associated human cases of salmonellosis. The pork industry is increasingly focused on the cost-effectiveness of surveillance while maintaining consumer confidence in the pork food supply. Using national control programme data from 2003 and 2004, we developed a zero-inflated binomial model to predict which farms were most at risk of Salmonella. We preferentially sampled these high-risk farms using two sampling schemes based on model predictions resulting from a farm’s covariate pattern and its random effect. Zero-inflated binomial modelling allows assessment of similarities and differences between factors that affect herd infection status (introduction), and those that affect the seroprevalence in infected herds (persistence and spread). Both large (producing greater than 5000 pigs per annum), and small herds (producing less than 2000 pigs per annum) were at significantly higher risk for infection and subsequent seroprevalence, when compared with medium sized herds (producing between 2000 and 5000 pigs per annum). When compared with herds being located elsewhere, being located in the south of Jutland significantly decreased the risk of herd infection, but increased the risk of a pig from an infected herd being seropositive. The model suggested that many of the herds where Salmonella was not detected were infected, but at a low prevalence. Using cost and sensitivity, we compared the results of our model based sampling schemes with those under the standard sampling scheme, based on herd size, and the recently introduced risk-based approach. Model-based results were less sensitive but show significant cost savings. Further model refinements, sampling schemes and the methods to evaluate their performance are important areas for future work, and these should continue to occur in direct consultation with Danish authorities.


Impacts

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References
  • • 
    Implementing reduction strategies for surveillance systems that were developed to protect public health is challenging. The apparently contradictory requirements for continued consumer confidence in food supply, and cost reduction for industry need to be met.
  • • 
    We use zero-inflated modelling to demonstrate how targeting farms in space and by risk factors has the potential to result in more efficient ways of conducting surveillance.
  • • 
    This allows assessment of the factors that affect herd infection status (introduction) and those that affect the sero-prevalence in infected herds (persistence and spread). Partitioning risk factors is valuable for directing practical risk-mitigating advice.

Introduction

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References

New challenges for animal health surveillance for zoonotic disease in the twenty-first century are many and include those brought about by increased trade, limited resources, consumer awareness and disease emergence (Hodges and Kimball, 2005; Woolhouse and Gowtage-Sequeria, 2005; Fevre et al., 2006; Vorou et al., 2007). This study is focused on the additional challenge of developing reduction strategies for surveillance systems for diseases that in the past represented an important risk, while today the risk to consumers is substantially reduced. Surveillance for Salmonella in Danish pig herds is an example that meets these criteria.

Such strategies require a delicate balance between satisfying producer and industry concerns about cost-effective testing and maintaining consumer confidence in food supply. Salmonella and BSE were the food risks most dreaded in a UK survey of food risk perception undertaken in 1999 (Kirk et al., 2002) and a recent survey of consumers identified meat as the food item in which confidence had decreased the most (Verbeke et al., 2007). It makes sense that any strategy involving a reduction in testing should demonstrate an equal or greater sensitivity as the existing one, regardless of the potential efficiency gains.

The means to evaluate the sensitivity of a surveillance programme and subsequently compare alternatives has been explored in the veterinary epidemiological literature (Audigé et al., 2001; Cannon, 2002; Martin et al., 2007b). In this context, a surveillance programme is considered as a diagnostic system which aims to correctly identify the presence or absence of an unwanted agent. By quantifying the characteristics of the diagnostic system (such as its specificity and sensitivity), a surveillance programme can be formally evaluated. For example, Audigé and Beckett defined surveillance sensitivity as the probability of declaring an area infected, given that infection exists, for the evaluation of surveillance for porcine reproductive and respiratory syndrome (PRRS) in Switzerland (Audigé and Beckett, 1999).

Quantification of the sensitivity of a surveillance system allows comparison of alternative surveillance strategies. For example, the comparison of the sensitivity of the currently targeted surveillance system for classical swine fever (CSF) in Denmark with that of a simulated non-targeted system identified that the current system was twice as sensitive compared with the simulated, non-targeted system (Martin et al., 2007a). In another Danish example, the sensitivity of the current surveillance programme for infectious bovine rhinotracheitis (IBR) was compared with three other surveillance scenarios targeting specific geographical areas and risk periods (Chriel et al., 2005).

Techniques all involving scenario tree methodology have been used for proof of disease freedom for exotic, non-zoonotic and clinically severe animal infections such as PRRS, CSF and IBR. In this study, we apply zero-inflated binomial modelling to the endemic, zoonotic and sub-clinical infection of Danish finisher pigs with Salmonella spp. using routinely collected data from the existing DSSCP data set from 2003 and 2004. Proof of disease freedom is not the end-point here, rather the issue is to maintain the status of domestically produced pork as a minor source of salmonellosis in humans. We propose it is possible to meet the differing needs of both consumer confidence in food supply and industry requirements for a surveillance reduction strategy with a targeted approach, whereby populations with higher risk of infection are preferentially sampled. These higher-risk populations are identified by model-based predictions driven by the previous year’s performance and their covariate risk profile. The objective was firstly to develop a model that predicts which farms are most at risk of Salmonella. Second, we preferentially sample the high-risk farms and compare our results to those under: (i) the standard sampling scheme, based on herd size and (ii) the recently introduced risk-based approach (Ministry of Family and Consumer Affairs, 2006). In this way, we are able to evaluate the impact of alternative sampling strategies on overall system performance.

Materials and Methods

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References

Data sources

Data were obtained from three sources. First, every pig herd is required to register with the Danish Central Husbandry Register. This provided a unique identifier (the CHR number), details of farm location, herd size and the number of sows in the herd.

The second source of data was from the central database of the DSSCP. We used the results from 9735 farms in 2003 (n = 578 260 individual samples) for initial model building. The DSSCP database also provided results from the 8151 farms sampled in 2004 that were also sampled in 2003 to investigate our different sampling schemes. Details retrieved from the DSSCP database included the CHR number, the date of sampling and the result of the Danish-mix ELISA (DME). This test measures antibodies in meat-juice to determine the previous exposure of finisher pigs to Salmonella spp. and can detect O-antigens from at least 93% of all serovars known to be present in Danish pigs (Mousing et al., 1997). The principal advantages of serological methods for Salmonella detection is the ability to assay a large number of samples rapidly at relatively low cost and high sensitivity when compared with bacteriology (2€ per sample). For these analyses, an ELISA optical density percentage (OD%) greater than 20 is classified as positive. This is equivalent to an adjusted OD% of greater than 10: the cut-off for positivity that has been used by the DSSCP since 1 August 2001 (Alban et al., 2002). All samples included in this study were analysed at the Danish Institute for Food and Veterinary Research using the DME. On the basis of testing, herds receive a monthly ‘serological Salmonella index’ which is based on a weighted average of the results from the previous three months. The levels of index are low level or no antibodies (index 0–39); medium (index 40–69); and high (index 70 or greater) (Alban et al., 2002). Herds in the medium and high index have reduced payments for finisher pigs sent to slaughter and must collect pen-faecal samples to determine the subtype and distribution of Salmonella in the herd.

The third source of data was the Danish Specific Pathogen Free (SPF) Company which provided health status details associated with each farm.

We chose to analyse data from 2003 and 2004 as we had access to additional farm-level details such as herd size, health status and the number of sows on the farm for those respective years. We proposed fitting a model to data from 2003 to inform sampling strategies for the subsequent year (estimation). Then we fit a model to the 2004 data and use this to see how successful the sampling strategies chosen from the 2003 data were (prediction).

Sampling schemes and model development

Four sampling schemes were developed. Herds were assigned to one of these schemes based on estimates from a model fitted to DSSCP data from 2003:

  • 1
    Original herd size based sampling (OHS): This sampling strategy was in place from August 2001 to July 2005. Using this approach, the eligible population comprised all herds with an annual kill greater than 200 slaughter pigs (representing 99% of all finisher herds in Denmark). The number of samples taken depended solely on herd size: the aim was to take 60, 75 or 100 samples annually from herds with an estimated annual kill of 200–2000, 2001–5000 and greater than 5000 slaughter pigs respectively (Alban et al., 2002). For the purposes of this study, we have used this sampling scheme to represent the bench-mark to which we compare the alternative sampling strategies.
  • 2
    DMA risk-based sampling (DRB): In July 2005, the surveillance system became performance-based which reduced the annual sample size by approximately one-third. For herds that had no positive meat juice samples over the previous 5 months, the sample size was reduced to one sample per month (Enoe et al., 2003; Ministry of Family and Consumer Affairs, 2006). If a herd then had one or more positive samples, the strategy reverts to one based on herd-size (OHS). We apply a modified version of these sampling criteria to herds in 2004 based on their performance in 2003. Our modification is that we have extended the time period over which herds are assessed to determine their prevalence to be the whole year, rather than the previous 5 months.
  • 3
    Model derived risk-based sampling A (MRBA): We developed a targeted surveillance strategy based on our previous risk-factor, spatial and temporal analyses of the DSSCP data (Benschop et al., 2008a, b, c). All herds with a predicted median within-herd seroprevalence at or below a model determined cut-off in 2003 were identified as low risk and were placed on the DRB scheme. This prediction was based on the farm’s covariate pattern and random farm effect. All other herds (above the predicted within-herd seroprevalence threshold) were left on the current sampling scheme for 2004 based on herd size (OHS).
  • 4
    Model derived risk-based sampling B (MRBB): As in MRBA above, all herds with a predicted median within-herd seroprevalence at or below a model determined cut-off in 2003 were identified as low risk and were placed on the DRB scheme. The remaining herds were then assigned to two different sampling schemes depending on their predicted seroprevalence in 2003: (i) those with a predicted seroprevalence that was <0 : 25 or >0 : 55 were left on the current sampling scheme based on herd size; and, (ii) those with a predicted seroprevalence of between 0.25 and 0.55 were more intensively sampled to provide 95% confidence that we were within 0.05 of the true value of the predicted seroprevalence. The increased intensity of sampling is created from model-derived data. This range was chosen as these herds were near the cut-off for level 2 Salmonella status (0.40).

Model development for the sampling schemes

The frequency histogram of the herd-level prevalence based on the actual test results from the OHS sampling strategy for 2003 and 2004 (Figure 1) showed a large amount of variation with a predominance of test-negative herds. These test-negative herds can come from two types of disease-negative herds: (i) those that are truly uninfected and therefore every sample is negative, and (ii); those that are, in fact, infected but provide insufficient samples to detect the presence of infection. This led us to propose a zeroinflated binomial (ZIB) approach to model herd-level Salmonella prevalence as it reflected our understanding of what is happening on the farm. The ZIB model has two herd-level outcomes, the probability of infection and — conditional on infection being present — an estimate of herd-level seroprevalence. This type of modelling can provide an added advantage over logistic regression: an ability to assess the extent of the similarities and differences between factors affecting herd infection status (invasion) and those affecting the seroprevalence in infected herds (persistence and spread).

image

Figure 1.  The distribution of actual within-herd prevalence for Salmonella in Danish pig herds, 2003. Data originate from the Danish swine Salmonella surveillance-and-control programme.

Download figure to PowerPoint

Variables that might explain both the presence of infection and herd-level prevalence included herd size, farm location, the number of sows present and herd health status. Herd size was the actual number of slaughter pigs produced for the year; this was centred by subtracting the mean and dividing by 1000. Farm location was a binary variable; if a herd was located in the Sonderjylland district it was coded as 1, otherwise 0. Health status was a three-level categorical variable: conventional, SPF and SPF with Mycoplasma. The presence of sows was expressed as a three level ordinal variable: farms with no sows, farms with less than 125 (some) and farms with over 125 (many).

Logistic regression modelling was used for initial model building. Bivariate analyses found all covariates significant at the P ≤ 0.25 level and using data from 2003 we built a multivariable model within the statistical software r, version 2.5.1 (Ihaka and Gentleman, 1996). The outcome variable was seroprevalence defined as the number of cases divided by the number of samples taken. All putative risk factors were significant. The continuous variable herd size was checked to see if it was linear in its log odds (Hosmer and Lemeshow, 1989). Polynomials of herd size and biologically plausible two-way interaction terms between the main-effect variables were considered for inclusion.

Once satisfied with the model structure we developed a logistic model within a Bayesian framework using winbugs version 1.4.1 (Gilks et al., 1994). The code for the model is shown in Fig. 2. Initially, we stipulated informed priors for the intercept term, and covariates relating to location, health status and the number of sows present on farm. We based these on published literature supplying subjective information about the likelihood ascribed to various combinations of covariate values (Congdon, 2001). For example, from earlier work on other data from the Danish Salmonella surveillance-and-control programme we believed that it would be protective factor for a herd having SPF health status (Benschop et al., 2008b). Moreover, residing in the district of Sonderjylland in the south of Jutland would be a risk factor (Benschop et al., 2008a) for herd-level sero-positivity. Based on available literature, an increased number of sows on farms were considered a risk factor for Salmonella in finishers (Hautekiet et al., 2008).

image

Figure 2. winbugs code for the zero-inflated binomial model.

Download figure to PowerPoint

Priors for the Bayesian logistic regression model were expressed in terms of a conjugate beta density (Congdon, 2001). We used a non-informed, normally distributed prior centred at zero and with a variance of 1 for the effect of herd size, given information about the effect of this variable on sero-positivity was not certain or conflicting. Three chains were run and convergence was judged to have occurred on the basis of visual inspection of time series plots and Gelman-Rubin plots (Toft et al., 2007). The length of the chain was determined by running sufficient iterations to ensure the Monte Carlo standard errors for each parameter were less than 5% of the posterior standard deviation. A total of 40 000 iterations were run with a ‘burn in’ of 4000 iterations.

The logistic regression model was extended to a zero-inflated binomial model and specified as follows:

  • image

Here, the number of cases from the ith herd is binomially distributed as a function of the number of trials (tests for Salmonella antibodies in meat-juice) pop[i], and the probability of a test being positive (adjusted OD% > 10), p[i].

We further defined:

  • image

where J[i] is an indicator variable representing infection status of the ith herd, rho[i] is the sero-prevalence conditional on the presence of infection. The term rho therefore represents the probability of finding infection in a randomly chosen pig from an infected herd. The latent variable J[i] is distributed as:

  • image

where q[i] is the probability of a herd being infected. This latent variable was modelled as:

  • image(1)

In Equation 1, the logit of the observed probability of the ith herd being infected, logit(qi), was modelled as a function of = 4 farm-level explanatory variables (herd size, location, the number of sows present and health status) and a random effect term, Ai, which was normally distributed with a mean of zero and precision σ. For the ZIB model, the continuous variable herd size was categorized to facilitate model convergence. The categories chosen were the same as those used in the DSSCP (Alban et al., 2002).

The latent variable rho[i] was modelled as:

  • image(2)

In Equation 2, the logit of the probability of observing infection in a randomly chosen pig from the ith infected farm was modelled as a function of the four farm-level explanatory variables defined earlier and a random effect term for herd, Bi, which was normally distributed with a mean of zero and precision τ.

We set non-informed, normally distributed priors centred at zero and with a precision of 0.5 for each of the fixed effect terms, including the intercept. Sensitivity to these priors was evaluated by re-running the models with a precision of 1 and 0.2. For the precision of the random farm-level effects, σ and τ, we specified a precision of 1. Sensitivity to these priors was evaluated by re-running the models with a precision of 0.5 and 0.3.

Three chains were run and convergence was judged to have occurred on the basis of visual inspection of plots of the sampled values as a time series (Toft et al., 2007). The required number of iterations of the Gibbs sampler was determined by running sufficient iterations to ensure the Monte Carlo standard errors for each parameter were less than 5% of the posterior standard deviations. A total of 30 060 iterations were run with a ‘burn in’ of 1000 iterations.

We proposed fitting this model on 2003 data to inform sampling strategies for the subsequent year (estimation). Then we fit a model to the 2004 data and use this to see how successful the sampling strategies chosen from the 2003 data were by, for example, comparing the number of false negatives (prediction).

To check for consistency between years (2003 and 2004), we examined model outputs from both years of data separately and compared the magnitude and direction of the regression coefficients. The 8151 random farm-level effects for the 2 years were compared using scatter-plots and quantified using Lin’s concordance correlation coefficient (Lin, 1989).

A scatter plot of the median conditional sero-prevalence rho[i] versus the median probability of infection q[i] (Fig. 3) was used to identify the cut-off for the two model derived risk-based sampling schemes MRBA and MRBB.

image

Figure 3.  Scatter plot of median predicted sero-prevalence as a function of median predicted probability of infection derived from a zero-inflated binomial model. Farms with at least one positive sample detected in are represented by the black dots, those with no positive samples detected are represented by the grey dots. Data are from 8151 farms sampled in the Danish swine Salmonella surveillance-and-control programme in 2003.

Download figure to PowerPoint

Comparison of sampling schemes

The results from all four sampling schemes were compared by considering cost, the number of false-negative farms and the number of farms detected with a within herd sero-prevalence of ≥0.40.

Costs were compared by adding up the number of tests taken under each of the four sampling schemes. Only the costs of meat juice testing were taken into account, with each meat juice sample tested costing 2€. These costs are borne by the producers through levies on each pig slaughtered. There are follow-on tests once herds reach level 2 and 3 of 200€ with further costs if herds are found to be positive. These follow-on tests were not considered further in this study.

For each farm (= 8151) there were 1020 iterations stored from the model and these were used to determine the false-negative rate and the number of farms detected with a within-herd sero-prevalence of ≥0.40 for each of the four sampling schemes.

The number of farms that were falsely reported as negative and the sensitivity for each of the four sampling schemes was determined using the following process:

  • (a) 
    the J[i] parameter, the indictor variable representing infection status of the ith herd, for 2004 was examined at each iteration. If it equalled one, then, for that iteration, the farm was considered infected. Otherwise, for that iteration, the farm was considered uninfected;
  • (b) 
    rho[i], the predicted within-herd seroprevalence given the herd was infected, for 2004 was determined for each iteration when the farm was infected. rho[i] was combined with the number of pigs sampled, using the binomial distribution to determine the number of positives that would be detected at each iteration;
  • (c) 
    a false-negative iteration was defined as one where the farm was infected at the iteration, but no positives were detected at that iteration. The number of false-negative iterations was summed and divided by the number of total iterations to give the number of false-negative farms;
  • (d) 
    this was expressed as the sensitivity of the sampling scheme by dividing the number of false-negative farms by the total number of farms (= 8151), and subtracting this fraction (the false-negative fraction) from one.

The number of farms that were predicted to have an observed seroprevalence of ≥0.40 for each of the four sampling schemes was determined using the following process:

  • 1
    the number of positives detected in each herd for each iteration was determined as in steps (a) and (b) in the preceding paragraph;
  • 2
    the number of positives was divided by the number sampled to give the observed seroprevalence in each herd at each iteration;
  • 3
    these numbers were summed and divided by the number of iterations to obtain the expected number of herds with observed seroprevalences of ≥0 : 40.

Results

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References

Data sources

In 2003, there were 9735 herds in the programme. The median number of pigs finished per year was 2000 (IQR: 800–3700). In total, 5938 herds (61%) kept no sows, 1752 (18%) some and 2045 (21%) kept many. A total of 7107 herds (73%) were of conventional health status, 586 (6%) SPF status and 2042 (21%) SPF with Mycoplasma. Finally, 978 herds (10%) were from Sonderjylland.

Sampling schemes (including model development)

All predictors were significant in the simple logistic regression model developed in r. The results of the Bayesian model using all these predictors are shown in Table 1. Compared with pigs from conventional health status herds, pigs from SPF health status and SPF-Mycoplasma status herds had 0.69 (95% CI 0.66–0.72) and 0.93 (95% CI 0.91–0.96) times the odds of being Salmonella positive, respectively. Compared with herds having 1–125 sows, having none or more than 125 sows increased the odds of a pig being Salmonella positive by a factor of 1.33 (95% CI 1.28–1.38) and 1.36 (95% CI 1.32–1.41), respectively. Compared with farms located outside of Sonderjylland, the odds of pigs being Salmonella positive on farms within Sonderjylland was increased by a factor of factor of 1.32 (95% CI 1.28–1.36).

Table 1.   Results of a logistic regression model showing factors associated with Salmonella sero-positivity in 578 260 meat-juice ELISA results taken from 9735 Danish finisher herds in 2003 as a part of the national surveillance-and-control programme
VariableLevelPosterior meanPosterior SDMC errorOR (95% CI)
  1. SD, standard deviation; CI, Bayesian credible interval; MC error, Monte Carlo standard error of the posterior mean; OR, odds ratio.

  2. aNo. finishers produced (rescaled by subtracting the minimum, then dividing by 1000).

  3. bInterpretation: once adjusted for herd size, sow status and location within Sonderjylland, a pig on a farm with SPF health status had 0.69 times the odds of being Salmonella positive compared with a pig on a farm with conventional health status (95%CI: 0.66–0.72).

Intercept−2.880.01<0.001
Herd sizeaContinuous2.9 × 10−20.00<0.0011.02 (1.01–1.03)
Health StatusConventionalReference
SPF−0.370.02<0.0010.69 (0.66–0.72)b
SPF (with Mycoplasma)−0.070.01<0.0010.93 (0.91–0.96)
Sow StatusNo sows0.310.02<0.0011.33 (1.28–1.38)
Some sows (1–125)Reference
Many sows (>125)0.280.02<0.0011.36 (1.32–1.41)
SonderjyllandNoReference
Yes0.280.02<0.0011.32 (1.28–1.36)

Estimated coefficients for the ZIB model are shown in Tables 2 and 3. Table 2 shows the factors included in the zero-inflated part of the model; these are interpreted as factors associated with the probability of a herd being infected. A herd producing less than 2000 (small), or greater than 5000 (large) pigs for slaughter per year had a 1.58 (95% CI: 1.18–2.11) or 2.08 (95% CI: 1.42–3.14) greater odds of infection with Salmonella, respectively, compared with herds producing between 2000 and 5000 (medium) pigs per year for slaughter. Compared with herds within farms located outside of Sonderjylland, the odds of a Sonderjylland herd being infected with Salmonella was decreased by a factor of 0.25 (95% CI: 0.19–0.33).

Table 2.   Zero-inflated binomial model output showing factors associated with Salmonella infection status in 8151 Danish finisher herds in 2003 as a part of the national surveillance and control programme
VariableLevelPosterior meanPosterior SDMC errorOR (95% CI)
  1. SD, standard deviation; CI, Bayesian credible interval; MC error, Monte Carlo standard error of the posterior mean; OR, odds ratio.

  2. aInterpretation: once adjusted for herd size, number of sows and herd health status, a farm located in Sonderjylland had 0.25 times the odds of being Salmonella positive compared with a farm located elsewhere (95%CI: 0.19–0.33).

Intercept2.360.190.008
Herd sizeSmall0.460.150.0041.58 (1.18–2.11)
MediumReference
Large0.730.210.0052.08 (1.42–3.14)
Health statusConventionalReference
SPF0.560.460.0111.67 (0.85–5.02)
SPF (with Mycoplasma)−0.140.160.0030.87 (0.63–1.18)
Sow statusNone0.140.210.0081.15 (0.75–1.71)
SomeReference
Many0.050.240.0091.04 (0.64–1.67)
SonderjyllandNoReference
Yes−1.380.140.0030.25 a (0.19–0.33)
Table 3.   Zero-inflated binomial model output showing factors associated with Salmonella seropositivity in 8151 Danish finisher herds in 2003 as a part of the national surveillance and control programme
VariableLevelPosterior meanPosterior SDMC errorOR (95% CI)
  1. SD, standard deviation; CI, Bayesian credible interval; MC error, Monte Carlo standard error of the posterior mean; OR, odds ratio.

  2. aInterpretation: once adjusted for herd size, number of sows and herd health status, a pig on a farm located in Sonderjylland had 1.68 times the odds of being Salmonella positive compared with a pig on a farm located elsewhere (95%CI: 1.51–1.86).

Intercept−3.330.040.002
Herd sizeaSmall0.150.040.0021.16 (1.08–1.24)
MediumReference
Large0.150.050.0021.16 (1.06–1.28)
Health StatusConventionalReference
SPF−0.370.070.0020.69 (0.61–0.78)
SPF (with Mycoplasma)−0.080.040.0010.92 (0.85–0.99)
Sow StatusNone0.280.050.0031.32 (1.20–1.45)
SomeReference
Many0.280.060.0031.33 (1.19–1.48)
SonderjyllandNoReference
Yes0.520.050.0021.68a (1.51–1.86)

Table 3 shows the model results for the binomial part of the ZIB model; these are interpreted as variables associated with the level of seropositivity in a herd, given that the herd is infected. The odds of a pig being sero-positive in an infected small or large herd was increased by a factor of 1.16 (95% CI 1.08–1.24) compared with a pig being sero-positive in an infected medium herd. The remaining results were similar to those provided in Table 1 for the logistic regression model.

The ZIB model was insensitive to changes in the precision parameter of the prior distribution assigned to Ai and Bi. The zero-inflated part of the model showed a 5-fold increase in the value of the posterior standard deviation when compared with the binomial part of the model.

As we planned to use this model, based on 2003 data, to predict the probability of infection and seropositivity in 2004 we checked for consistency between the 2 years. This was thought to be important, because substantial changes in pig- and herd-level risks for infection (arising from, e.g. changes in herd size or changes in the price of feed) from 1 year to the next could reduce the ability of the 2003 model to predict herd-level behaviour in 2004. The magnitude and sign of the regression coefficients for 2003 and 2004 were compared. There was no change in sign of the estimated regression coefficients for each year. The two alpha coefficients (for the variables SPF health status and many sows) showed minor changes in magnitude between years with overall conclusions remained unchanged. For example, the alpha coefficient for SPF health status changed from 0.56 in 2003 to 0.36 in 2004. The beta coefficients (for variables associated with sero-positivity, given infection) were similar between years.

There was moderate positive correlation between the random farms effects for the 2 years. The Lin’s concordance correlation coefficient for the random effect terms Ai from the zero-inflated part of the ZIB model for the 2 years was 0.18 (95% CI: 0.16–0.20), the location shift parameter was 0.004 and the scale-shift parameter was 0.89. The equivalent measures for the random effect terms Bi from the sero-positivity part of the ZIB model were 0.52 (95% CI: 0.51–0.54), −0.003 and 0.99 respectively.

A scatter plot of the median conditional sero-prevalence q[i] versus the median probability of infection rho[i] for 2003 in shown in Fig. 3. There is a partial distinction in predicted seroprevalence between farms that were detected as positive (red dots) and those that were not (green dots). This provided us with our cut-off for 0.09 for the sampling schemes. The decision to select this cut-point was subjectively based on visual appraisal of the plot. This plot suggests that many of the farms where the disease has not been observed are actually infected but at a low prevalence.

Comparison of sampling schemes

Table 4 shows the performance of each of the schemes. The scheme with the lowest cost was MRBA; the one with the highest cost was OHS. The one with the lowest number of false negatives and highest sensitivity was OHS and the one with the highest number of false negatives and lowest sensitivity was MRBA and MRBB. The scheme that reported the largest number of high-positive farms was MRBA and MRBB.

Table 4.   Performance of four sampling schemes for surveillance for Salmonella in Danish finisher herds in 2004, n = 8151 herds
Sampling schemeOHSDRBMRBAMRBA
  1. aFarms infected in 2004 but not detected by the sampling scheme.

  2. bFarms the sampling scheme has detected at a Salmonella seroprevalence of ≥0.40.

No. false-negative farmsa731118632573251
Sensitivity0.910.850.600.60
No. high-positive farmsb30484911481199
Cost of scheme (€1000)1118959372479

Discussion

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References

We report on the use of a zero-inflated binomial model to investigate the performance of alternative sampling strategies for zoonotic Salmonella. To the best of our knowledge, practical applications of this technique in the veterinary literature are scarce. Reports of the counterparts of these techniques for count data, zero-inflated Poisson and zero-inflated negative binomial models are more numerous. In Indonesia, Cheung (2006) used a ZIB model in the regression analysis of the cognitive function of Indonesian children in relation to weight-for-age in infancy and childhood. In an ecological context, Martin et al. (2005) discuss the use of zero-inflated binomial and Poisson models in situations where both true and false zeroes occur. They analyse site occupancy data on four woodland bird species in Southern Australia using habitat type and landscape metrics. As far as we are aware, ours is the first application of zero-inflated binomial modelling to endemic disease surveillance; there is potential to use this approach in the design of other surveillance systems.

Our model is based on our earlier work that found on-farm risk factors and broad spatial location to be associated with Salmonella status (Benschop et al., 2008a; b). Our proposed sampling strategies were not based on season as we found this was not an associated risk factor (Benschop et al., 2008c). However, it is important to consider this in a targeted surveillance strategy as time of year commonly has an effect on infectious disease, for example, the pattern of human cases of salmonellosis consistently reports a late summer–early autumn peak in Denmark (Hald and Andersen, 2001) and Canada (Guerin et al., 2005). This seasonality may be due to both direct and indirect effects of climate. The effect of season has modified surveillance strategies for infectious diseases. In Denmark, for example, it has been recommended that sampling for infectious bovine rhinotracheitis (IBR) occurs primarily during the winter months of the year (Chriel et al., 2005).

One variable, herd size, was a significant risk factor for both infection with Salmonella and subsequent sero-prevalence. When compared with medium size herds (producing 2000–5000 pigs for slaughter per annum) both large (producing greater than 5000 pigs) and small (producing less than 2000 pigs) size herds were at greater risk. Keeping none or greater than 125 sows was a risk factor when compared with keeping 1–125 in the sero-prevalence model only. It is likely the results for these two variables act through mechanisms such as buying in and mixing of pigs, number of visitors, biosecurity measures, feeding systems and other management factors (Leontides et al., 2003; Farzan et al., 2006).

When compared with being located elsewhere, being located in Sonderjylland significantly decreased the odds of a herd being infected with Salmonella but significantly increased the odds of a pig from an infected herd being Salmonella positive. There is no ready to hand explanation for this seeming paradox but it may be related to the herd demographics within this region. This region forms a border with Germany and there are two distinct types of farms present: small family owned more traditional operations and larger modern premises, particularly on the island of Als. Earlier work of ours has reported that farms in this region (compared with all other regions) showed the most variation in farm level prevalence of Salmonella (Benschop et al., 2009).

The use of the ZIB model has the potential to allow assessment of the extent of the similarities and differences between factors that affect herd infection status (introduction) and those that affect the sero-prevalence in infected herds (persistence and spread). One can think of invasion as things to do with external biosecurity such as rodent control number and type of suppliers, and visitor policy, whereas persistence and spread falls under internal biosecurity such as type of partitions between pens (Lo Fo Wong et al., 2004) and use of an all-in-all-out production system versus a continuous one (Beloeil et al., 2004; Lurette et al., 2008).

There is a significant amount of literature that has preceded this work in the determination of sampling strategy to where the DSSCP has evolved to today (Mousing et al., 1997; Nielsen et al., 2001; Alban et al., 2002; Ekeroth et al., 2003; Enoe et al., 2003). The challenge remains that if too few samples are taken then there is little chance of detecting a positive in an infected herd. Additionally, a small numbers of samples may make it too easy for an infected herd to reach a cut-off proportion purely by chance. The current risk-based system seems a reasonable compromise in that only 12 samples a year are taken from herds that have been consistently negative (Ministry of Family and Consumer Affairs, 2006). Once a positive is detected, herds return to the higher intensity herd-size based system with 60, 75 or 100 samples taken before a threshold criterion is applied. Our model supports that decision by providing reassurance that even if a non-detected farm is infected it will most likely be infected at a low sero-prevalence (Fig. 1). Owners are highly motivated to avoid entry into levels 2 or 3 within the higher intensity herd-size based system. If they do enter, they try to leave as soon as possible as there are considerable costs associated with these levels. This is one of the key reasons why predicting how a herd will perform based on the previous years’ performance is complex.

We present only a few sampling schemes as this study’s primary focus is the development of the concept rather than fine-tuning for the ‘best’ sampling scheme. As expected the scheme with the lowest cost (€372 000) was MRBA; the next lowest at almost one-third as much again was MRBB at €479 000. It is important to reiterate that for each scenario these are the direct costs associated only with sampling meat-juice at the abattoir; no follow on costs associated with on farm testing, hygienic slaughter or carcass downgrading are included. A full-cost benefit analysis is beyond the scope of this study but would be important groundwork prior to implementation of a change in the sampling regime.

The schemes with the lowest sensitivity were also MRBA and MRBB. We defined false negatives as farms that were infected in 2004 but had no positives detected by the sampling scheme. Both model-based sampling schemes had 39% of farms falling into this category. If the aim of the scheme is to detect every infected herd then these schemes perform poorly. However, the aim of the scheme is to identify herds with a high sero-prevalence, the detection of every infected herd is not an aim. In fact, our work would suggest that there are very few if any herds that are not infected and most herds (98%) have a predicted sero-prevalence below 40% (see Fig. 1).

We have fitted the model at one point in time, using accumulated data from 2003 to determine sampling for 2004. This model could be updated on a monthly basis allowing incorporation of the latest meat-juice results, and allowing for dynamic changes in covariates as herds increase in size, no longer keep sows, or enter an SPF health status, to take several examples. The development of model refinements, sampling schemes and the methods to evaluate their performance are important areas for future work and would make the best use of this new tool. This should continue to occur in direct consultation with Danish authorities.

References

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Materials and Methods
  6. Results
  7. Discussion
  8. References
  • Alban, L., H. Stege, and J. Dahl, 2002: The new classification system for slaughter-pig herds in the Danish Salmonella surveillance-and-control program. Prev. Vet. Med. 53, 133146.
  • Audigé, L., and S. Beckett, 1999: A quantitative assessment of the validity of animal-health surveys using stochastic modelling. Prev. Vet. Med. 38, 259276.
  • Audigé, L., M. G. Doherr, R. Hauser, and M. D. Salman, 2001: Stochastic modelling as a tool for planning animal-health surveys and interpreting screening-test results. Prev. Vet. Med. 49, 117.
  • Beloeil, P., P. Fravalo, C. Fablet, J. Jolly, E. Eveno, Y. Hascoet, C. Chauvin, G. Salvat, and F. Madec, 2004: Risk factors for Salmonella enterica subsp. enterica shedding by market-age pigs in French farrow-to-finish herds. Prev. Vet. Med. 63, 103120.
  • Benschop, J., M. L. Hazelton, M. A. Stevenson, J. Dahl, R. S. Morris, and N. P. French, 2008a: Descriptive spatial epidemiology of subclinical Salmonella infection in finisher pig herds: application of a novel method of spatially adaptive smoothing. Vet. Res. 39, 2. doi: 10.1051/vetres:2007040.
  • Benschop, J., M. A. Stevenson, J. Dahl, and N. P. French, 2008b: Towards incorporating spatial risk analysis for Salmonella sero-positivity into the Danish swine surveillance programme. Prev. Vet. Med. 83, 347359.
  • Benschop, J., M. A. Stevenson, J. Dahl, R. S. Morris, and N. P. French, 2008c: Temporal and longitudinal analysis of Danish Swine Salmonellosis Control Programme data: implications for surveillance. Epidemiol. Infect. 136, 15111520.
  • Benschop, J., M. A. Stevenson, J. D., R. S. Morris, and N. French, 2009: Informing surveillance programmes by investigating spatial dependency of subclinical Salmonella infection. Epidemiol. Infect. 137, 13481359.
  • Cannon, R. M., 2002: Demonstrating disease freedom-combining confidence levels. Prev. Vet. Med. 52, 227249.
  • Cheung, Y. B., 2006: Growth and cognitive function of Indonesian children: zero-inflated proportion models. Stat. Med. 25, 30113022.
  • Chriel, M., M. D. Salman, and B. A. Wagner, 2005: Evaluation of surveillance and sample collection methods to document freedom from infectious bovine rhinotracheitis in cattle populations. Am. J. Vet. Res. 66, 21492153.
  • Congdon, P., 2001: Bayesian Statistical Modelling, pp. 531. John Wiley & Sons, Chichester.
  • Ekeroth, L., L. Alban, and N. Feld, 2003: Single versus double testing of meat-juice samples for Salmonella antibodies, in the Danish pig-herd surveillance programme. Prev. Vet. Med. 60, 155165.
  • Enoe, C., H. Wachmann, and J. Boes 2003: Low intensity serological surveillance for Salmonella enterica in slaughter pigs from low prevalence herds in Denmark. In: Proceedings of the International Symposium of Veterinary Epidemiology and Economics, Vina del Mar, Chile.
  • Farzan, A., R. M. Friendship, C. E. Dewey, K. Warriner, C. Poppe, and K. Klotins, 2006: Prevalence of Salmonella spp. on Canadian pig farms using liquid or dry-feeding. Prev. Vet. Med. 73, 241254.
  • Fevre, E. M., B. M. Bronsvoort, K. A. Hamilton, and S. Cleaveland, 2006: Animal movements and the spread of infectious diseases. Trends Microbiol. 14, 125131 (Epub 2006 Feb 2007).
  • Gilks, W. R., A. Thomas, and D. J. Spiegelhalter, 1994: A language and program for complex Bayesian modelling. Stat. 43, 169178.
  • Guerin, M. T., S. W. Martin, and G. A. Darlington, 2005: Temporal clusters of Salmonella serovars in humans in Alberta, 1990–2001. Can. J. Public Health 96, 390395.
  • Hald, T., and J. S. Andersen, 2001: Trends and seasonal variations in the occurrence of Salmonella in pigs, pork and humans in Denmark, 1995–2000. Berl. Munch. Tierarztl. Wochenschr. 114, 346349.
  • Hautekiet, V., V. Geert, V. Marc, and G. Rony, 2008: Development of a sanitary risk index for Salmonella seroprevalence in Belgian pig farms. Prev. Vet. Med. 86, 75.
  • Hodges, J. R., and A. M. Kimball, 2005: The global diet: trade and novel infections. Global Health 1, 4.
  • Hosmer, D., and S. Lemeshow, 1989: Applied Logistic Regression, pp. 307. John Wiley and Sons Inc., New York.
  • Ihaka, R., and R. Gentleman, 1996: r: A Language for Data Analysis and Graphics. J Graph. Comput. Stat. 5, 299314.
  • Kirk, S. F., D. Greenwood, J. E. Cade, and A. D. Pearman, 2002: Public perception of a range of potential food risks in the United Kingdom. Appetite 38, 189197.
  • Leontides, L. S., E. Grafanakis, and C. Genigeorgis, 2003: Factors associated with the serological prevalence of Salmonella enterica in Greek finishing swineherds. Epidemiol. Infect. 131, 599606.
  • Lin, L. I., 1989: A concordance correlation coefficient to evaluate reproducibility. Biometrics 45, 255268.
  • Lo Fo Wong, D., J. Dahl, H. Stege, P. Van Der Wolf, L. Leontides, A. Von Altrock, and B. Thorberg, 2004: Herd-level risk factors for subclinical Salmonella infection in European finishing-pig herds. Prev. Vet. Med. 62, 253266.
  • Lurette, A., C. Belloc, S. Touzeau, T. Hoch, P. Ezanno, H. Seegers, and C. Fourichon, 2008: Modelling Salmonella spread within a farrow-to-finish pig herd. Vet. Res. 39, 49 (Epub 2008 Jun 2014).
  • Martin, T., B. Wintle, J. Rhodes, P. Kuhnert, S. Field, S. Low-Choy, A. Tyre, and H. Possingham, 2005: Zero tolerance ecology: improving ecological inference by modelling the source of zero observations. Ecol. Lett. 8, 12351246.
  • Martin, P. A., A. R. Cameron, K. Barfod, E. S. Sergeant, and M. Greiner, 2007a: Demonstrating freedom from disease using multiple complex data sources 2: case study--classical swine fever in Denmark. Prev. Vet. Med. 79, 98115 (Epub 2007 Jan 2018).
  • Martin, P. A., A. R. Cameron, and M. Greiner, 2007b: Demonstrating freedom from disease using multiple complex data sources 1: a new methodology based on scenario trees. Prev. Vet. Med. 79, 7197 (Epub 2007 Jan 2016).
  • Ministry of Family and Consumer Affairs, 2006: Annual Report on Zoonoses in Denmark 2005. The Danish Zoonoses Centre, Copenhagen.
  • Mousing, J., P. T. Jensen, C. Halgaard, F. Bager, N. Feld, B. Nielsen, J. P. Nielsen, and S. Bech-Nielsen, 1997: Nation-wide Salmonella enterica surveillance and control in Danish slaughter swine herds. Prev. Vet. Med. 29, 247261.
  • Nielsen, B., L. Alban, H. Stege, L. Sorensen, V. Mogelmose, J. Bagger, J. Dahl, and D. Baggesen, 2001: A new Salmonella surveillance and control programme in Danish pig herds and slaughterhouses. Berl. Munch. Tierarztl. Wochenschr. 114, 323326.
  • Toft, N., G. T. Innocent, G. Gettinby, and S. W. Reid, 2007: Assessing the convergence of Markov Chain Monte Carlo methods: an example from evaluation of diagnostic tests in absence of a gold standard. Prev. Vet. Med. 79, 244256.
  • Verbeke, W., L. J. Frewer, J. Scholderer, and H. F. De Brabander, 2007: Why consumers behave as they do with respect to food safety and risk information. Anal. Chim. Acta 586, 27 (Epub 2006 Aug 2001).
  • Vorou, R. M., V. G. Papavassiliou, and S. Tsiodras, 2007: Emerging zoonoses and vector-borne infections affecting humans in Europe. Epidemiol. Infect. 135, 12311247.
  • Woolhouse, M. E., and S. Gowtage-Sequeria, 2005: Host range and emerging and reemerging pathogens. Emerg. Infect. Dis. 11, 18421847.